Re: Large CVS Installations

[Rachel Burns]

A great way of handling redundancy and scaling is to
use WANdisco CVS Replicator (http://www.wandisco.com/cvs).

With it you can setup shared-nothing architecture.  All the
CVS repositories/replicas run on independent machines (heterogeneous
ok)  but are clustered together  using the WANdisco replicator.

Advantages are reads can scale horizontally, as most CVS traffic
is typically generated by reads (co, up, log, stat etc) you can immediately
offload disk, RAM and CPU subsystems.  Writes (commits, tag) are replicated
using a Quorum based approach which allows any node to be written to,
no single point of failure or bottleneck unlike traditional single 
master schemes. 

Basically you get the benefits of multiple servers processing power but 
still keep the
virtual repository abstraction - true clustering.

Redundancy and High availability is also a side effect of the shared-nothing
architecture.

DHARNA, AJAY [AG/1000] wrote:

> Hi all,
>
> I am trying to get a general idea on how you handle redundancy and scaling of 
> Large CVS installations. Does anyone deal with a large enough CVS installation 
> to tell me how you are managing to handle this?
>
> I would like to know whether they have spread the repository over multiple 
> servers to handle a large number of users and whether you think that this is 
> good or bad. Whether they are using clustering to do this and again what your 
> opinions are on this. (Good and bad)
>
> I would like to know whether you have a 1 disk volume mounted to multiple CVS 
> server machines or whether you prefer to do it is another way and why you think 
> that that method is more efficient.
>
> Thank you.
>
> Ajay Dharna
>
>
> _______________________________________________
> Info-cvs mailing list
> Info-cvs@gnu.org
> http://lists.gnu.org/mailman/listinfo/info-cvs
>
>